Blacklist all functions involved while handling a trap. We:
- convert some of the symbols into private symbols, and
- blacklist most functions involved while handling a trap.
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
It is actually safe to probe system_call() in entry_64.S, but only till
we unset MSR_RI. To allow this, add a new symbol system_call_exit()
after the mtmsrd and blacklist that.
Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
It is common to get a PMU interrupt right after the mtmsr instruction that
enables interrupts. Due to this, the stack trace profile gets needlessly split
across system_call_common() and system_call().
Previously, system_call() symbol was at the current place to hide a few
earlier symbols which have since been made private or removed entirely.
So, let's move system_call() slightly higher up, right after the mtmsr
instruction that enables interrupts. Convert existing references to
system_call to a local syscall symbol.
Suggested-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Convert some of the symbols into private symbols and blacklist
system_call_common() and system_call() from kprobes. We can't take a
trap at parts of these functions as either MSR_RI is unset or the
kernel stack pointer is not yet setup.
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
[mpe: Don't convert system_call_common to _GLOBAL()]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Commit b48bbb82e2 ("powerpc/64s: Don't unbalance the return branch
predictor in __replay_interrupt()") introduced __replay_interrupt_return
symbol with '.L' prefix in hopes of keeping it private. However, due to
the use of LOAD_REG_ADDR(), the assembler kept this symbol visible. Fix
the same by instead using the local label '1'.
Fixes: Commit b48bbb82e2 ("powerpc/64s: Don't unbalance the return branch
predictor in __replay_interrupt()")
Suggested-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Merge our fixes branch, a few of them are tripping people up while
working on top of next, and we also have a dependency between the CXL
fixes and new CXL code we want to merge into next.
Use the different spin loop primitives in some simple powerpc
spin loops, including those which will spin as a common case.
This will help to test the spin loop primitives before more
conversions are done.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Add some includes of <linux/processor.h>]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
pnv_wakeup_noloss() expects r12 to contain SRR1 value to determine if the wakeup
reason is an HMI in CHECK_HMI_INTERRUPT.
When we wakeup with ESL=0, SRR1 will not contain the wakeup reason, so there is
no point setting r12 to SRR1.
However, we don't set r12 at all so r12 contains garbage (likely a kernel
pointer), and is still used to check HMI assuming that it contained SRR1. This
causes the OPAL msglog to be filled with the following print:
HMI: Received HMI interrupt: HMER = 0x0040000000000000
This patch clears r12 after waking up from stop with ESL=EC=0, so that we don't
accidentally enter the HMI handler in pnv_wakeup_noloss() if the value of
r12[42:45] corresponds to HMI as wakeup reason.
Prior to commit 9d29250136 ("powerpc/64s/idle: Avoid SRR usage in idle
sleep/wake paths") this bug existed, in that we would incorrectly look at SRR1
to check for a HMI when SRR1 didn't contain a wakeup reason. However the SRR1
value would just happen to never have bits 42:45 set.
Fixes: 9d29250136 ("powerpc/64s/idle: Avoid SRR usage in idle sleep/wake paths")
Signed-off-by: Akshay Adiga <akshay.adiga@linux.vnet.ibm.com>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Change log and comment massaging]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
nr_cpu_ids can be limited by nr_cpus boot parameter, whereas NR_CPUS is a
compile time constant, which shouldn't be compared against during cpu kick.
Signed-off-by: Santosh Sivaraj <santosh@fossix.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
During secondary start, we do not need to BUG_ON if an invalid CPU number
is passed. We already print an error if secondary cannot be started, so
just return an error instead.
Signed-off-by: Santosh Sivaraj <santosh@fossix.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Around 95% of memory is reserved by fadump/capture kernel. All this
memory is freed, one page at a time, on writing '1' to the node
/sys/kernel/fadump_release_mem. On systems with large memory, this
can take a long time to complete, leading to soft lockup warning
messages. To avoid this, add reschedule points at regular intervals.
Also, while memblock_reserve() implicitly takes care of holes in the
given memory range while reserving memory, those holes need to be
taken care of while releasing memory as memory is freed one page at
a time. Add support to skip holes while releasing memory.
Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Hari Bathini <hbathini@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
fadump fails to register when there are holes in boot memory area.
Provide a helpful error message to the user in such case.
Signed-off-by: Hari Bathini <hbathini@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
To register fadump, boot memory area - the size of low memory chunk that
is required for a kernel to boot successfully when booted with restricted
memory, is assumed to have no holes. But this memory area is currently
not protected from hot-remove operations. So, fadump could fail to
re-register after a memory hot-remove operation, if memory is removed
from boot memory area. To avoid this, ensure that memory from boot
memory area is not hot-removed when fadump is registered.
Signed-off-by: Hari Bathini <hbathini@linux.vnet.ibm.com>
Reviewed-by: Mahesh J Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
fadump sets up crash memory ranges to be used for creating PT_LOAD
program headers in elfcore header. Memory chunk RMA_START through
boot memory area size is added as the first memory range because
firmware, at the time of crash, moves this memory chunk to different
location specified during fadump registration making it necessary to
create a separate program header for it with the correct offset.
This memory chunk is skipped while setting up the remaining memory
ranges. But currently, there is possibility that some of this memory
may have duplicate entries like when it is hot-removed and added
again. Ensure that no two memory ranges represent the same memory.
When 5 lmbs are hot-removed and then hot-plugged before registering
fadump, here is how the program headers in /proc/vmcore exported by
fadump look like
without this change:
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
NOTE 0x0000000000010000 0x0000000000000000 0x0000000000000000
0x0000000000001894 0x0000000000001894 0
LOAD 0x0000000000021020 0xc000000000000000 0x0000000000000000
0x0000000040000000 0x0000000040000000 RWE 0
LOAD 0x0000000040031020 0xc000000000000000 0x0000000000000000
0x0000000010000000 0x0000000010000000 RWE 0
LOAD 0x0000000050040000 0xc000000010000000 0x0000000010000000
0x0000000050000000 0x0000000050000000 RWE 0
LOAD 0x00000000a0040000 0xc000000060000000 0x0000000060000000
0x000000019ffe0000 0x000000019ffe0000 RWE 0
and with this change:
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
NOTE 0x0000000000010000 0x0000000000000000 0x0000000000000000
0x0000000000001894 0x0000000000001894 0
LOAD 0x0000000000021020 0xc000000000000000 0x0000000000000000
0x0000000040000000 0x0000000040000000 RWE 0
LOAD 0x0000000040030000 0xc000000040000000 0x0000000040000000
0x0000000020000000 0x0000000020000000 RWE 0
LOAD 0x0000000060030000 0xc000000060000000 0x0000000060000000
0x000000019ffe0000 0x000000019ffe0000 RWE 0
Signed-off-by: Hari Bathini <hbathini@linux.vnet.ibm.com>
Reviewed-by: Mahesh J Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
On POWER9 the ERAT may be incorrect on wakeup from some stop states
that lose state. This causes random segvs and illegal instructions
when these stop states are enabled.
This patch invalidates the ERAT on wakeup on POWER9 to prevent this
from causing a problem.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Merge comment change with upstream changes]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The asm code assumes the FP regs are at the start of fp_state. While
this is true now, it may not always be the case and there is nothing
enforcing it.
This fixes the asm-offsets to point to the actual FP registers inside
the fp_state. Similarly for VMX.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The P9 PVR bits 12-15 don't indicate a revision but instead different
chip configurations. From BookIV we have:
Bits Configuration
0 : Scale out 12 cores
1 : Scale out 24 cores
2 : Scale up 12 cores
3 : Scale up 24 cores
DD1 doesn't use this but DD2 does. Linux will mostly use the "Scale
out 24 core" configuration (ie. SMT4 not SMT8) which results in a PVR
of 0x004e1200. The reported revision in /proc/cpuinfo is hence
reported incorrectly as "18.0".
This patch fixes this to mask off only the relevant bits for the major
revision (ie. bits 8-11) for POWER9.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Emergency stacks have their thread_info mostly uninitialised, which in
particular means garbage preempt_count values.
Emergency stack code runs with interrupts disabled entirely, and is
used very rarely, so this has been unnoticed so far. It was found by a
proposed new powerpc watchdog that takes a soft-NMI directly from the
masked_interrupt handler and using the emergency stack. That crashed
at BUG_ON(in_nmi()) in nmi_enter(). preempt_count()s were found to be
garbage.
To fix this, zero the entire THREAD_SIZE allocation, and initialize
the thread_info.
Cc: stable@vger.kernel.org
Reported-by: Abdul Haleem <abdhalee@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Move it all into setup_64.c, use a function not a macro. Fix
crashes on Cell by setting preempt_count to 0 not HARDIRQ_OFFSET]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This converts the powerpc VDSO time update function to use the new
interface introduced in commit 576094b7f0 ("time: Introduce new
GENERIC_TIME_VSYSCALL", 2012-09-11). Where the old interface gave
us the time as of the last update in seconds and whole nanoseconds,
with the new interface we get the nanoseconds part effectively in
a binary fixed-point format with tk->tkr_mono.shift bits to the
right of the binary point.
With the old interface, the fractional nanoseconds got truncated,
meaning that the value returned by the VDSO clock_gettime function
would have about 1ns of jitter in it compared to the value computed
by the generic timekeeping code in the kernel.
The powerpc VDSO time functions (clock_gettime and gettimeofday)
already work in units of 2^-32 seconds, or 0.23283 ns, because that
makes it simple to split the result into seconds and fractional
seconds, and represent the fractional seconds in either microseconds
or nanoseconds. This is good enough accuracy for now, so this patch
avoids changing how the VDSO works or the interface in the VDSO data
page.
This patch converts the powerpc update_vsyscall_old to be called
update_vsyscall and use the new interface. We convert the fractional
second to units of 2^-32 seconds without truncating to whole nanoseconds.
(There is still a conversion to whole nanoseconds for any legacy users
of the vdso_data/systemcfg stamp_xtime field.)
In addition, this improves the accuracy of the computation of tb_to_xs
for those systems with high-frequency timebase clocks (>= 268.5 MHz)
by doing the right shift in two parts, one before the multiplication and
one after, rather than doing the right shift before the multiplication.
(We can't do all of the right shift after the multiplication unless we
use 128-bit arithmetic.)
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Acked-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Since trace_clock is in a different file and already marked with notrace,
enable tracing in time.c by removing it from the disabled list in Makefile.
Also annotate clocksource read functions and sched_clock with notrace.
Testing: Timer and ftrace selftests run with different trace clocks.
Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Santosh Sivaraj <santosh@fossix.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
As for slb_miss_realmode(), rename slb_allocate_realmode() to avoid
confusion over whether it runs in real or virtual mode - it runs in
both.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
slb_miss_realmode() doesn't always runs in real mode, which is what the
name implies. So rename it to avoid confusing people.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
All the callers of slb_miss_realmode currently open code the #ifndef
CONFIG_RELOCATABLE check and the branch via CTR in the RELOCATABLE case.
We have a macro to do this, BRANCH_TO_COMMON(), so use it.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
The SLB miss handler uses r3 for the faulting address but r12 is
mostly able to be freed up to save r3 in. It just requires SRR1
be reloaded again on error.
It would be more conventional to use r12 for SRR1 (and use r11 to
save r3), but slb_allocate_realmode clobbers r11 and not r12.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The EXCEPTION_PROLOG_1 used by SLB miss already saves CTR when the
kernel is built with CONFIG_RELOCATABLE. So it does not have to be
saved and reloaded when branching to slb_miss_realmode. It can be
restored from the PACA as usual.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The EX_DAR save area is only used in exceptional cases. With r3 no
longer clobbered by slb_allocate_realmode, saving faulting address to
EX_DAR can be deferred to those cases.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
In a busy system, idle wakeups can be expected from IPIs and device
interrupts.
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Idle code now always runs at the 0xc... effective address whether
in real or virtual mode. This means rfid can be ditched, along
with a lot of SRR manipulations.
In the wakeup path, carry SRR1 around in r12. Use mtmsrd to change
MSR states as required.
This also balances the return prediction for the idle call, by
doing blr rather than rfid to return to the idle caller.
On POWER9, 2-process context switch on different cores, with snooze
disabled, increases performance by 2%.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Incorporate v2 fixes from Nick]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Have the system reset idle wakeup handlers branched to in real mode
with the 0xc... kernel address applied. This allows simplifications of
avoiding rfid when switching to virtual mode in the wakeup handler.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The __replay_interrupt() code is branched to with bl, but the caller is
returned to directly with rfid from the interrupt.
Instead, rfid to a stub that returns to the caller with blr, which
should keep the return branch predictor balanced.
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
msgsnd doorbell exceptions are cleared when the doorbell interrupt is
taken. However if a doorbell exception causes a system reset interrupt
wake from power saving state, the message is not cleared. Processing
the doorbell from the system reset interrupt requires msgclr to avoid
taking the exception again.
Testing this plus the previous wakup direct patch gives:
original wakeup direct msgclr
Different threads, same core: 315k/s 264k/s 345k/s
Different cores: 235k/s 242k/s 242k/s
Net speedup is +10% for same core, and +3% for different core.
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
When the CPU wakes from low power state, it begins at the system reset
interrupt with the exception that caused the wakeup encoded in SRR1.
Today, powernv idle wakeup ignores the wakeup reason (except a special
case for HMI), and the regular interrupt corresponding to the
exception will fire after the idle wakeup exits.
Change this to replay the interrupt from the idle wakeup before
interrupts are hard-enabled.
Test on POWER8 of context_switch selftests benchmark with polling idle
disabled (e.g., always nap, giving cross-CPU IPIs) gives the following
results:
original wakeup direct
Different threads, same core: 315k/s 264k/s
Different cores: 235k/s 242k/s
There is a slowdown for doorbell IPI (same core) case because system
reset wakeup does not clear the message and the doorbell interrupt
fires again needlessly.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This simplifies the asm and fixes irq-off tracing over sleep
instructions.
Also move powersave_nap check for POWER8 into C code, and move
PSSCR register value calculation for POWER9 into C.
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
On Power9, trying to use data breakpoints throws the splat shown
below. This is because the check for a data breakpoint in DSISR is in
do_hash_page(), which is not called when in Radix mode.
Unable to handle kernel paging request for data at address 0xc000000000e19218
Faulting instruction address: 0xc0000000001155e8
cpu 0x0: Vector: 300 (Data Access) at [c0000000ef1e7b20]
pc: c0000000001155e8: find_pid_ns+0x48/0xe0
lr: c000000000116ac4: find_task_by_vpid+0x44/0x90
sp: c0000000ef1e7da0
msr: 9000000000009033
dar: c000000000e19218
dsisr: 400000
Move the check to handle_page_fault() so as to catch data breakpoints
in both Hash and Radix MMU modes.
We have to change the check in do_hash_page() against 0xa410 to use
0xa450, so as to include the value of (DSISR_DABRMATCH << 16).
There are two sites that call handle_page_fault() when in Radix, both
already pass DSISR in r4.
Fixes: caca285e5a ("powerpc/mm/radix: Use STD_MMU_64 to properly isolate hash related code")
Cc: stable@vger.kernel.org # v4.7+
Reported-by: Shriya R. Kulkarni <shriykul@in.ibm.com>
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
[mpe: Fix the fall-through case on hash, we need to reload DSISR]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
ftrace_caller() depends on a modified regs->nip to detect if a certain
function has been livepatched. However, with KPROBES_ON_FTRACE, it is
possible for regs->nip to have been modified by the kprobes pre_handler
(jprobes, for instance). In this case, we do not want to invoke the
livepatch_handler so as not to consume the livepatch stack.
To distinguish between the two (kprobes and livepatch), we check if
there is an active kprobe on the current function. If there is, then we
know for sure that it must have modified the NIP as we don't support
livepatching a kprobe'd function. In this case, we simply skip the
livepatch_handler and branch to the new NIP. Otherwise, the
livepatch_handler is invoked.
Fixes: ead514d5fb ("powerpc/kprobes: Add support for KPROBES_ON_FTRACE")
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
For DYNAMIC_FTRACE_WITH_REGS, we should be passing-in the original set
of registers in pt_regs, to capture the state _before_ ftrace_caller.
However, we are instead passing the stack pointer *after* allocating a
stack frame in ftrace_caller. Fix this by saving the proper value of r1
in pt_regs. Also, use SAVE_10GPRS() to simplify the code.
Fixes: 153086644f ("powerpc/ftrace: Add support for -mprofile-kernel ftrace ABI")
Cc: stable@vger.kernel.org # v4.6+
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This fixes a crash when function_graph and jprobes are used together.
This is essentially commit 237d28db03 ("ftrace/jprobes/x86: Fix
conflict between jprobes and function graph tracing"), but for powerpc.
Jprobes breaks function_graph tracing since the jprobe hook needs to use
jprobe_return(), which never returns back to the hook, but instead to
the original jprobe'd function. The solution is to momentarily pause
function_graph tracing before invoking the jprobe hook and re-enable it
when returning back to the original jprobe'd function.
Fixes: 6794c78243 ("powerpc64: port of the function graph tracer")
Cc: stable@vger.kernel.org # v2.6.30+
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Acked-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The ISA v3.0B copy-paste facility only requires cpabort when switching
to a process that has foreign real addresses mapped (direct access to
accelerators), to clear a potential copy buffer filled by a previous
thread. There is no accelerator driver implemented yet, so cpabort can
be removed. It can be be re-added when a driver is implemented.
POWER9 DD1 requires the copy buffer to always be cleared on context
switch, but if accelerators are not in use, then an unpaired copy from
a dummy region is sufficient to clear data out of the copy buffer.
This increases context switch performance by about 5% on POWER9.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The sync (aka. hwsync, aka. heavyweight sync) in the context switch
code to prevent MMIO access being reordered from the point of view of
a single process if it gets migrated to a different CPU is not
required because there is an hwsync performed earlier in the context
switch path.
Comment this so it's clear enough if anything changes on the scheduler
or the powerpc sides. Remove the hwsync from _switch.
This improves context switch performance by 2-3% on POWER8.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
There is no need to explicitly break the reservation in _switch,
because we are guaranteed that the context switch path will include a
larx/stcx.
Comment the guarantee and remove the reservation clear from _switch.
This is worth 1-2% in context switch performance.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Commit 4387e9ff25 ("[POWERPC] Fix PMU + soft interrupt disable bug")
hard disabled interrupts over the low level context switch, because
the SLB management can't cope with a PMU interrupt accesing the stack
in that window.
Radix based kernel mapping does not use the SLB so it does not require
interrupts hard disabled here.
This is worth 1-2% in context switch performance on POWER9.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The syscall exit code that branches to restore_math is quite heavy on
Book3S, consisting of 2 mtmsr instructions. Threads that don't use both
FP and vector can get caught here if the kernel ever uses FP or vector.
Lazy-FP/vec context switching also trips this case.
So check for lazy FP and vector before switching RI for restore_math.
Move most of this case out of line.
For threads that do want to restore math registers, the MSR switches are
still suboptimal. Future direction may be to use a soft-RI bit to avoid
MSR switches in kernel (similar to soft-EE), but for now at least the
no-restore
POWER9 context switch rate increases by about 5% due to sched_yield(2)
return performance. I haven't constructed a test to measure the syscall
cost.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
After bc3551257a ("powerpc/64: Allow for relocation-on interrupts from
guest to host"), a getppid() system call goes from 307 cycles to 358
cycles (+17%) on POWER8. This is due significantly to the scratch SPR
used by the hypercall check.
It turns out there are a some volatile registers common to both system
call and hypercall (in particular, r12, cr0, ctr), which can be used to
avoid the SPR and some other overheads. This brings getppid to 320 cycles
(+4%).
Testing hcall entry performance by running "sc 1" in guest userspace
before this patch is 854 cycles, afterwards is 826. Also a small win
there.
POWER9 syscall is improved by about the same amount, hcall not tested.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Supporting 512TB requires us to do a order 3 allocation for level 1 page
table (pgd). This results in page allocation failures with certain workloads.
For now limit 4k linux page size config to 64TB.
Fixes: f6eedbba7a ("powerpc/mm/hash: Increase VA range to 128TB")
Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
In commit 8c27226119 ("powerpc/numa: Enable USE_PERCPU_NUMA_NODE_ID"), we
switched to the generic implementation of cpu_to_node(), which uses a percpu
variable to hold the NUMA node for each CPU.
Unfortunately we neglected to notice that we use cpu_to_node() in the allocation
of our percpu areas, leading to a chicken and egg problem. In practice what
happens is when we are setting up the percpu areas, cpu_to_node() reports that
all CPUs are on node 0, so we allocate all percpu areas on node 0.
This is visible in the dmesg output, as all pcpu allocs being in group 0:
pcpu-alloc: [0] 00 01 02 03 [0] 04 05 06 07
pcpu-alloc: [0] 08 09 10 11 [0] 12 13 14 15
pcpu-alloc: [0] 16 17 18 19 [0] 20 21 22 23
pcpu-alloc: [0] 24 25 26 27 [0] 28 29 30 31
pcpu-alloc: [0] 32 33 34 35 [0] 36 37 38 39
pcpu-alloc: [0] 40 41 42 43 [0] 44 45 46 47
To fix it we need an early_cpu_to_node() which can run prior to percpu being
setup. We already have the numa_cpu_lookup_table we can use, so just plumb it
in. With the patch dmesg output shows two groups, 0 and 1:
pcpu-alloc: [0] 00 01 02 03 [0] 04 05 06 07
pcpu-alloc: [0] 08 09 10 11 [0] 12 13 14 15
pcpu-alloc: [0] 16 17 18 19 [0] 20 21 22 23
pcpu-alloc: [1] 24 25 26 27 [1] 28 29 30 31
pcpu-alloc: [1] 32 33 34 35 [1] 36 37 38 39
pcpu-alloc: [1] 40 41 42 43 [1] 44 45 46 47
We can also check the data_offset in the paca of various CPUs, with the fix we
see:
CPU 0: data_offset = 0x0ffe8b0000
CPU 24: data_offset = 0x1ffe5b0000
And we can see from dmesg that CPU 24 has an allocation on node 1:
node 0: [mem 0x0000000000000000-0x0000000fffffffff]
node 1: [mem 0x0000001000000000-0x0000001fffffffff]
Cc: stable@vger.kernel.org # v3.16+
Fixes: 8c27226119 ("powerpc/numa: Enable USE_PERCPU_NUMA_NODE_ID")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The i-side 0111b machine check, which is "Instruction Fetch to foreign
address space", was missed by 7b9f71f974 ("powerpc/64s: POWER9 machine
check handler").
The POWER9 processor core considers host real addresses with a
nonzero value in RA(8:12) as foreign address space, accessible only
by the copy and paste instructions. The copy and paste instruction
pair can be used to invoke the Nest accelerators via the Virtual
Accelerator Switchboard (VAS).
It is an error for any regular load/store or ifetch to go to a foreign
addresses. When relocation is on, this causes an MMU exception. When
relocation is off, a machine check exception. It is possible to trigger
this machine check by branching to a foreign address with MSR[IR]=0.
Fixes: 7b9f71f974 ("powerpc/64s: POWER9 machine check handler")
Reported-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Currently tsk->thread.load_tm is not initialized in the task creation
and can contain garbage on a new task.
This is an undesired behaviour, since it affects the timing to enable
and disable the transactional memory laziness (disabling and enabling
the MSR TM bit, which affects TM reclaim and recheckpoint in the
scheduling process).
Fixes: 5d176f751e ("powerpc: tm: Enable transactional memory (TM) lazily for userspace")
Cc: stable@vger.kernel.org # v4.9+
Signed-off-by: Breno Leitao <leitao@debian.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Currently tsk->thread->load_vec and load_fp are not initialized during
task creation, which can lead to garbage values in these variables (non-zero
values).
These variables will be checked later in restore_math() to validate if the
FP and vector registers are being utilized. Since these values might be
non-zero, the restore_math() will continue to save the FP and vectors even if
they were never utilized by the userspace application. load_fp and load_vec
counters will then overflow (they wrap at 255) and the FP and Altivec will be
finally disabled, but before that condition is reached (counter overflow)
several context switches will have restored FP and vector registers without
need, causing a performance degradation.
Fixes: 70fe3d980f ("powerpc: Restore FPU/VEC/VSX if previously used")
Cc: stable@vger.kernel.org # v4.6+
Signed-off-by: Breno Leitao <leitao@debian.org>
Signed-off-by: Gustavo Romero <gusbromero@gmail.com>
Acked-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Naveen N. Rao